HACKER Q&A
📣 albertdessaint

Why Isn't Anyone Building a Global Brain for LLMs?


LLMs thrive on high-quality data. While features like web search, file uploads, and integrations with diverse data sources are useful, there seems to be no system for sharing and accessing information between LLM users.

Take my case as an example: I lend money to businesses. Ideally, if someone asks ChatGPT a question like "How can I get B2B financing?" I’d want them to discover me and then be able to ask follow-up questions about financing options, eligibility criteria, required documents, and the process. From there, I could step in and provide personalized assistance.

While web search exists, the internet wasn’t built for the LLM era, and appearing in search results isn’t guaranteed. I’ve also thought about creating an AI agent, but discoverability remains the issue.

Wouldn't a new information system tailored for LLMs be valuable? Imagine an open ecosystem where structured and unstructured data can be shared and accessed by LLMs on behalf of users, powering apps and tools that unlock opportunities and save time.

Am I missing something here, or does this idea have merit?

  👤 toomuchtodo Accepted Answer ✓
How would you build this (which costs both infra and human time to build and operate for the collection of knowledge) to only have every commercial entity soak it up for private gain?

It would be swank to have something like Wikipedia, Stack Overflow, etc where you could compound returns on teaching the system, but someone has to pay for it, and lots of folks don't want their work (free or otherwise) to just go into Big Tech or AI Startup valuations and returns. Novel training data is precious, and democratization of generative AI outcomes is not a function of time, but of effort and resources.

👤 jonahbenton
Look for instance at MCP servers

https://github.com/modelcontextprotocol/servers

Commercial price/deal servers I suspect are just around the corner.

👤 codingdave
The question is for whom that idea has merit. For sales and marketing folks, yeah, absolutely they would love to be able to just push their marketing spend towards something that would buy LLM users' eyeballs. For everyone else, this sounds like just a new iteration of unwanted ads in our face, and would probably breed the same level of resentment, annoyance, and ultimately blockers to avoid it all.
👤 cpfiffer
Love this! I wrote a post just recently you might like:

Show HN: Comind – A cognitive layer for the ATProtocol/Bluesky social network | https://news.ycombinator.com/item?id=42975350

👤 ansonhw
I'm confused why search + LLM wouldn't work for this?